AI is making developers code faster — but most companies still aren't shipping faster
GitLab's 2026 AI Accountability Report found 78% of developers code faster with AI, but 79% say overall software delivery hasn't accelerated because review and governance — not writing code — are now the bottleneck.
7 July 2026
GitLab’s 2026 AI Accountability Report, based on a survey of over 1,500 developers and engineering leaders across six countries, found a split that a lot of commissioning conversations still miss: 78% of developers say they write code faster with AI tools. Only a minority say their organisation is actually delivering software faster overall.
Why the gap exists
The report’s own explanation is blunt: teams adopted AI coding tools faster than they built the governance to control them. 80% of respondents say tool adoption outpaced governance. 92% report governance problems with AI-generated code in production. Fewer than half can reliably trace which code was AI-written versus human-written, which tools touched a given file, or establish clean code provenance when something breaks.
In other words, AI compressed the writing step of software development and left the reviewing, testing, and trusting step exactly where it was — so the bottleneck just moved downstream. A pull request that took an hour to write and now takes ten minutes still needs the same review, the same test coverage, and the same sign-off before it’s safe to ship. If that review process hasn’t changed, the time saved on writing gets eaten by a queue at review.
So what
If you’re commissioning a build and a proposal leads with “we use AI tools, so we’ll be faster,” ask the follow-up question GitLab’s data suggests most teams can’t answer: how do you know which code was AI-generated, and what’s the review process for it? An agency that can answer that concretely — with defined review gates, provenance tracking, and testing that scales with output rather than staying fixed — is one where AI speed actually reaches your release date instead of stalling at code review. That’s the standard we hold our own AI-assisted development process to. Get in touch if you want to talk through what that looks like for your project.